French Roast: Consumer Response to International Conflict

526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
French Roast: Consumer Response to International Conflict - Evidence from Supermarket
Scanner Data∗
Sonal S. Pandya† and Rajkumar Venkatesan
This Draft: January 2015
Review of Economics and Statistics (forthcoming)
∗
For their feedback and suggestions we thank the editor, referees, Sean Carr, Christina
Davis, Peter Debaere, James Harrigan, David Leblang, Ariell Reshef, Sheetal Sekhri, Robert
Urbatsch, Hans-Joachim Voth, Frank Warnock, and seminar participants at Berkeley, the
Darden School of Business, and the UVa Quantitative Collaborative. Sarah Andrews, Adam
Hughes, Ellie Kaknes, Steven Liao, Brenton Peterson, and especially Selin Kesebir provided
outstanding research assistance. The Behavioral Research at Darden Lab provided valuable
research support. Pandya gratefully acknowledges the financial support of the Bankard Fund
for Political Economy.
†
Corresponding Author. Department of Politics, University of Virginia, P.O. Box 400787,
Charlottesville, VA, 22904-4787, Phone: (434) 243-1573, Fax: (434) 243-3359, Email:
[email protected]
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
Abstract
Do consumers boycott in response to international conflict? We show that during the 2003
US-France dispute over the Iraq War the market share of French-sounding, US supermarket brands declined. The dispute was a negative shock to US consumers’ associations with
France. French-sounding brands, which consumers perceive to be French imports but are
not, allow us to isolate the dispute’s effect on economic behavior, as these brands’ only link to
France is through consumers’ associations. Our estimates, derived from a nationwide sample
of weekly supermarket sales for over 8,000 brands, are robust to a variety of alternate explanations. Additionally, we show that supermarkets with a higher proportion of customers
who are US citizens (i.e. who more strongly identify with the US national identity) exhibited
sharper boycotts.
JEL Classification: D03, D12, F51, F52
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
I
Introduction
Does international conflict influence consumer behavior? Interstate disputes, from armed
conflicts to diplomatic disagreements, often provoke animosity toward foreign adversaries.1
But we lack consensus on whether this type of animosity produces observable changes in
consumer behavior.2 Boycott participation can be a purely expressive act to signal anger and
reinforce personal values (Sen et al 2001). Alternately, consumers may be unwilling to make
the sacrifices associated with the boycott (John and Klein 2003). This puzzle remains in large
part because we lack precise measures of boycott participation. Most studies infer consumer
response from indirect measures such as bilateral trade patterns (Glick and Taylor 2010,
Fuchs and Klann 2013), abnormal stock market returns (Fisman et al 2013), or consumer
surveys, which are typically inconsistent with actual behavior (Morowitz and Schmittlein
1992).
In this article, we show that during the 2003 US-France dispute over the invasion of Iraq,
US consumers reduced their purchases of non-French, French-sounding supermarket brands.
The dispute was an exogenous, negative shock to US consumers’ associations with France.
Our identification strategy exploits the existence of French-sounding brands that consumers
incorrectly perceive to be French imports. These brand names cue French nationality by
incorporating quintessentially French letter clusters (e.g. “oui” or “eau”) and diacritics such
as an accent mark (é) or circumflex (ô). Consumers need not be fluent in French to detect
the French cues in brands such as TRESemmé shampoo and Raison D’Être beer.3 Brands
marketed to appear French but not necessarily imported from France provide a clean measure
of consumer sentiment because they are not otherwise exposed to the negative shock.
In early February 2003, the US pressed for a United Nations (UN)-authorized invasion
of Iraq. France vocally opposed the invasion and threatened to use its veto power as a
permanent member of the UN Security Council. Figure I illustrates the sharp decline in
public opinion towards France over the course of the conflict. In the week ending February
9, 2003, the week before France announced its formal opposition to the invasion, 64% of
1
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
American survey respondents reported a favorable opinion of France. By the week ending
March 2, the middle of the dispute, only 27% maintained a favorable opinion. Widespread,
anger-fueled calls for Americans to repudiate the consumption of French products culminated
on March 11 when the US House of Representatives renamed French fries freedom fries in
all House-run eating establishments.
We exploit this exogenous shock in US attitudes towards France to estimate the effect of
Americans’ perceptions of France on their choice of grocery brands. For each week of 2003,
we analyze same-store changes in brands’ market shares as compared to the same week one
year prior in more than 1,100 US supermarkets. For each store, we observe sales of up
to 8,000 brands across 27 product categories. For each brand-product category-store-week,
our unit of observation, we model change in market share relative to the previous year.
We measure the strength of consumers’ associations with France through surveys in which
respondents indicate the country they associate with each brand.
Our identifying assumption is that the dispute influenced American supermarket purchases only via negative associations with France. The shock did not affect product quality,
supply, or other relevant dimensions of consumer behavior.4 By estimating the change in
same-store market share over the previous year, we hold constant time-invariant factors including variation across stores in the ex ante supply and demand for French-sounding brands.
We control for store-level price changes in the previous week, the only real-time response
available to retailers.
We estimate a difference-in-differences model and show that sales of French-sounding
brands declined in weeks with more intense media coverage of the US-France dispute. At a
key juncture in the dispute, the week ending March 16, brands that consumers most strongly
perceive as French saw a 0.40% average decline in store market share as compared to the
same week in 2002 after controlling for changes in brand price and number of varieties.
The magnitude of annual change is striking given that the average price discount generates
only a 4% increase in brand sales over the previous week (Assmus et al 1984, Ataman et al
2
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
2010). We perform a wide range of robustness tests: controls for pro-France sentiment, sales
of brands associated with other countries, fluctuations in consumers’ budgets, and public
opinion regarding the dispute. Additionally, we find no change among French-owned US
registered trademarks, or brands owned by companies headquartered in France.
Finally, we consider the precise mechanisms through which the dispute may have influenced consumers’ brand choices. We hypothesize the dispute represented an external
threat that magnified in/out-group distinctions, making nationality more salient in individual decision-making (Gennaioli and Shleifer 2010, Bordalo et al 2013, Kőszegi and Szeidl
2013). We test this claim by analyzing cross-sectional variation in brand-share fluctuations
across supermarkets for the week ending March 16, a week with particularly high anti-France
sentiment, and we show stores with a higher proportion of US citizens in their customer base
exhibit a greater decline in French-sounding brand sales. We verify customers’ household
income, partisanship, and affiliation with the military do not account for this decline.
Our study contributes to empirical research on boycott participation during international
conflict. Supermarket purchases are a frequent, consistent, and nearly universal form of
consumption directly influenced by attitudes towards foreign countries. Not only are these
data highly disaggregated, they also directly link economic behavior to the cues consumers
receive when making consumption choices. Given the wide divergence between perceived
and actual country of origin, we provide an accurate measure of consumer attitudes toward
specific countries.
As a metric of boycott participation, supermarket sales improve upon existing measures
in a few ways. Existing studies of the 2003 US-France conflict analyzing bilateral trade
(Michaels and Zhi 2010, Davis and Meunier 2011) and US sales of French wine (Aschenfelter
et al 2007, Chavis and Leslie 2009) have contradictory findings. By contrast, our sample
spans 27 categories of widely consumed grocery products. The most similar extant studies
examine sales of a single highly specialized product, such as automobiles (Hong et al 2011,
Fouka and Voth 2013) or wine. The high frequency of these data captures nuanced, short-
3
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
lived shifts in sentiment obscured in more aggregate measures.
The remainder of this article is organized into four sections. Section II briefly reviews
the dispute and the emergence of consumer animosity toward France. Section III describes
our measures of key concepts including product sales, perceived French nationality, and
anti-French sentiment. Section IV outlines our estimation strategy, baseline findings, and
results of robustness tests. Section V concludes by discussing the consequences of consumer
animosity for global market integration and political behavior, as well as the study’s broader
implications for economic interdependence and international conflict.
II
Timeline of 2003 US-France Conflict
In early 2003, France opposed the US-led invasion of Iraq and refused to support a UN
Security Council resolution endorsing it. France was not the only US ally to oppose the
invasion, but as a permanent member of the UN Security Council, its position was decisive.
Great Britain, another permanent member, was allied with the US, while the remaining two,
China and Russia, sided with France in pressing for additional UN weapons inspections in
lieu of an invasion.
Tensions began to escalate in late January as the US prepared the formal case for invasion.
French president Jacques Chirac declared France’s opposition to military action, pledging to
build an anti-war coalition among its European allies. On February 4, US Secretary of State
Colin Powell appeared before the UN to offer evidence Iraq was developing weapons of mass
destruction and had ties to the terrorist organization Al-Qaeda. In the following days, France
publicly questioned the validity of this evidence and reiterated its support for continued
UN weapons inspections. US-France relations deteriorated sharply on February 14 when
French Foreign Minister Dominque de Villepin, speaking to the UN, issued France’s strongest
statements to date against the invasion. In the following weeks, the dispute continued to
escalate, and on March 5, France formally announced it would veto any UN Security Council
resolution endorsing the invasion.5 On March 20, the US-led invasion of Iraq began.
4
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
Both US public opinion and media coverage of the dispute demonstrate a consistent
pattern of anti-France sentiment. Figure I summarizes US public opinion. Average support
for the invasion increased marginally as the dispute escalated, reaching its peak the week of
the invasion. Declining attitudes toward France track the timeline of the diplomatic dispute
across three sentiment measures. Attitudes toward France appear to have been more volatile
than general attitudes towards the invasion.
The earliest media reports of the dispute were in mid-February within days of Villepin’s
UN speech. Much discussion centered around the boycott of French products. Fox News
commentator Bill O’Reilly described the boycott as retribution against France. On March
10, O’Reilly addressed viewers:
“Talking Points” is about to say good-bye to Roquefort dressing, farewell to Louis
Vuitton, au revoir to Yves Saint Laurent. Sorry, guys. We’re going to miss you.
And say farewell to Pierre Cardin, while you’re at it. France has now hurt the
USA, and for many of us, payback time has arrived.
The top panel of Figure II plots the number of times Fox News reported on the dispute
each week. The dashed line is the number of programs on which anti-French sentiment was
expressed. The solid line is the total number of programs that report on the dispute, regardless of tone. We observe notable spikes in both total Fox News coverage and anti-France
sentiment coinciding with key turning points in mid-February and mid-March. Michaels
and Zhi (2010, 262) show a similar pattern in weekly US newspaper mentions of the phrase
“anti-French sentiment.”
On March 11, the boycott received de facto US government endorsement. Two Republican members of the US House of Representatives mandated that all of the body’s food
service outlets replace the menu items French fries and French toast with freedom fries and
freedom toast.6
Although public sentiment toward France was overwhelmingly negative, some Americans
did not support the boycott. CNN senior political analyst Bill Schneider said on March 16:
5
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
So, will we boycott French dressing and Russian dressing? We asked people what
they thought about the idea of calling French fries “freedom fries” which is what
they do in the House of Representatives. The public’s view: it is not a sincere
expression of patriotism.
The bottom panel of Figure II plots weekly CNN coverage of the dispute. It makes clear
media coverage of anti-France sentiment far exceeded that of pro-France sentiment. In a
robustness test, we assess the relative importance of anti- and pro-France sentiment for the
sale of French-sounding brands.
III
Data
Our central claim is that in 2003, American consumers boycotted brands they perceived
to be French. Our identification strategy rests on the assumption that the US-France conflict
was an exogenous shock to American consumers’ purchasing decisions and that the shock
influenced brands’ market share exclusively through its effects on consumers’ changed associations with France. Our tests require three types of information: weekly supermarket sales
data, a measure of perceived brand nationality, and measures of weekly sentiment toward
France.
III.A
Supermarket Scanner Data
We measure consumer response to the US-France conflict using weekly supermarket sales
data supplied by Information Resources Inc. (IRI), a leading source of US supermarket
scanner data (Bronnenberg, Kreuger, and Mela 2008). These data cover a representative
sample of 1,145 supermarkets across 50 IRI-designated geographic markets.
7
Figure III
maps the geographic coverage of our data. The 135 supermarket chains represented in the
data collectively account for 79% of US supermarket sales in 2003 (Market Share Reporter
2004).
6
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
We construct our store-level measure of consumer response using weekly unit sales for
8,644 brands across 27 categories of grocery products. Major supermarket chains stock
mature brands and maintain a relatively stable portfolio of brands within each store. We
aggregate data across multiple stock keeping unit (SKU) codes of a single brand-product
category (e.g. six-pack of Coke, two-liter bottle of Coke) but not across distinct but related
brands (e.g. Coke and Diet Coke). In addition to unit sales, our data reports price and size
of selection, which we use as control variables.
For each brand-product category-store in our dataset, we model the weekly change in
market-share annual growth rate. Our outcome of interest is indexed by:
i: 8,644 brands,
j: 1,154 supermarkets,
k: 27 product categories, and
t: 52 weeks.
A brand’s weekly store market share is the number of brand product units sold as a
percentage of all units in the product category sold in that store-week. For example, if
brand i in product category k (e.g. yogurt) had a 0.5% market share in a given store j for
week t, the brand accounted for half of all units of yogurt sold in that store in that week.
For every brand-product category-store-week in our sample, we calculate the change in
market share between 2002 and 2003 (Share03 -02ijkt ). Measuring a brand’s market share,
as opposed to the total number of units sold, allows us to scale that store’s sales of a
brand relative to overall demand for that product category in that store-week. Changes in
market share also capture shifts in demand for brands distinct from changes in demand for
a particular product category.
Measuring annual change in demand within each store allows us to hold constant all
time-invariant baseline characteristics of the store’s customer base that influence sales, especially ex ante customer preferences. If we were to observe sales only in 2003, we could not
differentiate between a change in demand and pre-existing low demand. For each store, we
7
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
retain only brands that were sold in all weeks of 2002 and 2003 so our results are not biased
by attrition nor entry. We also hold constant seasonal fluctuations in brands’ market share
by comparing shares to the same week in the prior year.
We generate our dependent variable by calculating the weekly change (between week t
and week t − 1) in annual market-share growth rate:
∆Share03 -02ijkt = Share03 -02ijkt − Share03 -02ijkt−1 .
Taking the weekly difference in annual share growth controls for variation across product
categories in purchase frequency. For instance, consumers typically purchase shampoo less
frequently than salty snacks. Weekly difference also controls for any systematic correlation
between the propensities to consume a particular product category and participate in the
boycott. By estimating a model of weekly change, we control for unit roots that may arise
with inclusion of lagged growth rates.
III.B
Perceived Brand Nationality
We measure perceived brand nationality using the product brand names supplied in our
sales data. We rely on brand name to indicate nationality because it is a highly salient,
readily available cue (Usunier and Shaner 2002).8 For American consumers, brand names
based on foreign languages frequently evoke associations with a foreign country through
distinctive letter combinations and special characters, such as umlauts and accent marks that
do not occur in English. Survey and experimental evidence shows consumers systematically
misidentify the national origin of products because they infer nationality from marketing
cues, rather than searching for country of origin labels (Samiee et al 2005; Balabanis and
Diamantopoulos 2011).
Brand nationality is a cue that operates outside of consumers’ conscious awareness in a
manner analogous to social stereotypes (Liu and Johnson 2005, Martin et al 2011). Consumers draw inferences based on prior associations between the implied country and the
product. A French-sounding brand name, for instance, cues “a rich network of associations
8
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
related to aesthetic sensitivity, refined taste, and sensory pleasure” (LeClerc et al 1994,
264-268).
We administered surveys to assess the perceived nationality of brands via Amazon.com’s
Mechanical Turk service (MTurk), an online marketplace for repetitive human coding tasks
paid by piece rate. Our survey presented respondents with the brand name of a product
and its product category and asked them to select the most relevant from a list of brand
nationalities. Seven respondents independently coded each brand.
A possible measurement concern is that we administered surveys in 2011 to approximate consumer perceptions in 2003. In the intervening years, marketing strategies may
have changed respondents’ brand associations. However, the basic feature of nationality
marketing would not have changed owing to the stickiness of basic brand characteristics.
Nationality perceptions could have also changed because of the 2003 dispute. American
companies, including French’s mustard, issued press releases to clarify the national origin of
their products. To the extent these efforts systematically changed consumer perceptions, we
underestimate the perceived French nationality of some brands in 2003, biasing against our
expected results.
F renchScorei takes values between 0 and 7 corresponding to the number of respondents
who deemed brand i to be French. Table I provides examples of brands at each variable value.
Brands with F renchScorei = 7 exhibit strong French nationality cues, such as an accent
mark (L’oréal), French words, or geographic references (Sans Sucre de Paris). Lower-scoring
brands have distinctively French elements but leave more doubt.
We demonstrate the disconnect between brands’ perceived and objective French origin
by comparing F renchScorei to two measures: whether the brand’s US trademark filing
indicates a French owner (F renchT rademarki ) and whether the company that owns the
brand is headquartered in France (F renchHQi ). We identify a brand to be objectively
French if its trademark filing shows the owner has a French postal address.9 To the best of
our knowledge, no other data sources systematically link brands to owners. The US Patent
9
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
and Trademark Office database includes current and expired filings such that we can match
brands to their trademark owner in 2003 and includes both public and private companies.
This measure omits trademarks registered to US subsidiaries of French corporations but
it uniquely captures the nationality of French-owned trademarks that are licensed to US
corporations.10 Not all brands in our data are registered trademarks, but the trademark filing
process is not biased against French-owned brands, so selection into the group of trademarked
brands should not correlate with a brand’s perceived nationality. We use trademark filings
to link brands to their corporate owners and then identify which owners are subsidiaries
of France-based companies using company information from two comprehensive corporate
databases: Thomas Reuters’ Worldscope and Lexis-Nexis’ company database. F renchHQi ’s
key virtue is that it detects French nationality that would be obscured when a non-French
subsidiary of a France-based company is the trademark’s legal owner.
Table II demonstrates the substantial gap between perceived and actual French origin.
The table disaggregates brands within each value of F renchScorei by various objective
metrics. Of the 37 brands all survey respondents perceived to be French (F renchScorei = 7),
France-headquartered companies own only four and only three report a French postal address
on their US trademark filing. At every value of F renchScorei , many other brands appear to
be French than have an objective connection to France. In our sample of 8,644 brand names,
852, or nearly 10% of the sample, have a F renchScorei ≥ 3. By comparison, a total of
27 brands, less than 0.004% of our sample, are French-owned trademarks. Of those Frenchowned trademarks, 18 also received a F renchScorei ≥ 3.11 France-headquartered companies
own 50 of the brands in our sample. This measure includes 31 brands the trademarkbased measure does not capture, 25 of which are razor blade brands trademarked to the US
subsidiary of the France-headquartered company Société Bic.12
In our empirical analysis, we confirm no change in the sales in objectively French brands
regardless of which measure we use.
10
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
III.C
Animosity toward France
We measure the intensity of anti-France sentiment each week by the number of Fox News
programs that reported on the US-France dispute. We collected data using transcripts of
all Fox News programs available in the Lexis-Nexis database.
13
News coverage may have
a causal effect on consumers’ propensity to boycott, but we cannot verify this outside of
an experimental setting.14 Media coverage tends to influence mass sentiment in the early
stages of foreign policy crises (Baum and Groeling 2010), but political ideology strongly
influences consumers’ choice of news sources (Iyengar and Hahn 2009) such that the propensity to participate in the boycott and watch Fox News are likely correlated. We follow many
mass media and politics studies in using Fox News to represent partisan political sentiment
(Groseclose and Milyo 2005, Gentzkow and Shapiro 2006).
We construct two versions of this variable, F oxDisputeM entionst , the weekly total number of Fox News programs that mention the dispute, and F oxAntiF ranceM entionst , the
subset of programs that reported anti-France sentiment, typically the boycott of French
products. The top panel of Figure II plots both variables for each week of 2003 and shows
the majority of Fox News coverage was about anti-France sentiment. Only one Fox News
program expressed pro-France sentiment in 2003. Our measure emphasizes the frequency of
coverage because brand switching is more likely with repetition of new information (Abraham
and Lodish 1990).
We model market-share changes as a function of media coverage in the same week. Studies
from diverse, relevant contexts - news coverage on food product health effects (Niederdeppe
and Frosch 2009), product placement in television programs (Russell 2002), and campaign
advertising (Gerber et al 2011, Mitchell 2012) - show any correlation with behavior dissipates
within a week. We confirm our results are robust to lagging the media variable by a week
and taking the natural log in case of diminishing marginal effects.
11
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
IV
Empirical Model and Results
IV.A
Outcome of Interest: Change in Brand Market Share
We estimate a difference-in-differences ordinary least squares model of weekly changes in
each brand’s rate of market-share growth (∆Share03 -02ijkt ):
∆Share03 -02ijkt = β1 F oxAntiF ranceM entionst + β2 F renchScorei + β3 ∆Price03 -02ijkt +
β4 ∆NumVariants03 -02ijkt + β5 ∆Share03 -02ijkt +
β6 F oxAntiF ranceM entionst ∗ F renchScorei + ijkt−1
where
∆Share03 -02ijkt
=
difference in share growth from 2002 to 2003
between week t and t − 1 for brand i-product
category k in store j,
F oxAntiF ranceM entionst
=
the number of Fox News programs in week
t that express anti-France sentiment in
connection with the US-France dispute,
F renchScorei
=
number of survey participants that deemed
brand i to be French,
∆Price03 -02ijkt
=
difference in price growth from 2002 to 2003
between week t and t − 1 for brand i-product
category k in store j,
∆NumVariants03 -02ijkt
=
difference in number of variants from 2002 to
2003 between week t and t − 1 for brand
i-product category k in store j,
∆Share03 -02ijkt−1
=
difference in share growth from 2002 to
2003 between week t − 2 and t − 3 for brand
i-product category k in store j, and
ijkt
=
normally distributed random error term.
12
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
As is standard in empirical marketing analyses, we control for three time-varying brandstore characteristics that influence fluctuations in market share (Ataman, Van Heerde, and
Mela 2010). ∆Price03 -02ijkt controls for exogenous price changes and the effect of promotional, time-limited price discounts.15 Non-pricing responses, such as advertising, were less
likely because they require longer lead times to implement. Price promotions are retailers’
fastest response to negative demand shocks.16 Retailers’ contracts with manufacturers forbid changes to products’ shelf space allocation and location, so no retailer-driven change in
product supply or location is possible.17
We also control for weekly changes in the number of varieties of a brand a store stocks in
a product category. All else equal, consumers are more likely to purchase a brand if a store
stocks more varieties. ∆NumVariants03 -02ijkt is the annual change in the share of brand
i-product category k’s product line length in store j from a year prior in week t.
On average, brand shares changed little from 2002 to 2003. Our controls for prices and
number of product varieties stocked were similarly stable, as is characteristic of sales in wellestablished grocery retailers.18 Finally, we include lagged market share, ∆Share03 -02ijkt−1 ,
to control for unobserved factors that affect product sales beyond price and product line
length.
IV.B
Baseline Model Results
Table III summarizes our baseline model results. Column 1 estimates are based on weekly
anti-France Fox news mentions (F renchScorei ∗F oxAntiF ranceM entionst ). The coefficient
on the interaction term F renchScorei ∗ F oxAntiF ranceM entionst is the estimated causal
effect of anti-France sentiment on sales of French-sounding brands. For each additional antiFrance Fox News mention in a week, the change in the growth rate of French brands’ market
share declines 0.005 percent points on average. Column 2 estimates are based on total Fox
mentions of the dispute, both neutral and anti-French, and the corresponding decline is 0.001
percent points for each additional mention.
13
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
Figure IV plots the predicted weekly change in growth rate, calculated from the Table
III, column 2 estimates for F renchScorei = 3 (moderately French) and F renchScorei = 7
(highly French). On average, each additional mention correlates with an approximately
0.067 percent point decline for highly French brands and a 0.037 percent point decline for
moderately French brands. The most anti-France Fox mentions were in the weeks ending
on April 20 (nine) and March 16 (eight) and the predicted decline in growth rate for highly
French brands was 0.36 percent points and 0.30 percent points respectively. The corresponding decline for moderately French brands was 0.12 percent points and 0.16 percent points
respectively.19
Given that we control for major drivers of consumption, including price and product
variety, the robust, statistically significant finding is surprising. The magnitude of financial
cost, however, is relatively modest. For the week ending March 16, we estimate the boycott’s
one-week financial cost to French brands in excess of $37 million.20 The magnitude of lost
sales is comparable to that of food brands receiving negative news coverage (Niederdeppe
and Frosch 2009). As Figure IV indicates, the decline in French-sounding brands’ market
share did not persist.
The coefficient on F renchScorei describes change in brand share growth in weeks without
anti-France sentiment. In both models, the positive and statistically significant coefficient indicates seemingly French brands see growth in market share over the previous year. In Model
2, based on F oxDisputeM entionst , the components of the interaction term are statistically
significant individually.
The controls perform as expected. Change in brand share growth declines with an increase in price and rises when additional product varieties are introduced. Consistent with
stylized facts in marketing research, both controls have substantively large effects on growthrate fluctuations (Ataman, Van Heerde, and Mela 2010). The positive coefficient for the
lagged dependent variable indicates weekly change correlates with fluctuations in the previous week.
14
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
IV.C
Robustness Tests
In many robustness tests, the estimated effect of the dispute on the sale of Frenchsounding brands is typically at least as large as in the baseline specification.
We examine whether countervailing pro-France sentiment influenced sales. Critics derided the boycott, and some opponents of the Iraq invasion advocated efforts at a counterboycott to increase the purchase of French products. We take CNN’s coverage of the dispute
as an indicator of overall pro-France sentiment. In 2003, CNN had only two anti-France
mentions; all others were pro-France or neutral. We construct two versions of this measure:
CN N DisputeM entionst , the total number of CNN programs that mention the dispute and
CN N P roF ranceM entionst , the number of pro-France and/or anti-boycott CNN mentions.
Estimates of the interaction of each of these variables with F renchScorei represent the causal
effect of the dispute on sales of seemingly French consumer products.
Table IV reports these expanded models’ estimates. The interaction term estimates in
column 1, based on total mentions for each network, show total CNN mentions had no
statistically significant correlation with fluctuations in French brand sales. The column 2
estimates reveal that for each pro-France mention in a week, the annual growth rate of
French-branded products was 0.002 percent points greater than that of non-French brands.
In this model, Fox News coverage corresponds to a larger decline in sales compared to the
baseline model. These findings suggest pro-France sentiment moderated the observed fall in
market share. The higher frequency of dispute coverage on Fox News is associated with a
net overall growth rate decline of highly French brands in the lead up to the Iraq war.
We also consider the dispute’s effect on other nationalities’ brands. The observed decline in French-sounding brands’ market shares could reflect something besides anti-French
sentiment, either consumer ethnocentrism, a rejection of all seemingly foreign brands (Klein
2002), or a stronger preference for brands that seem to be imports from US allies. Distinguishing between these explanations is necessary to specify the nuances of consumer response
to international conflict. We perform two placebo tests to disentangle these motives. First,
15
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
we examine the effect of anti-French sentiment on the sale of German-sounding brands. Germany allied with France in opposition to the invasion, but unlike France, Germany is not
a permanent member of the UN Security Council. News coverage of the dispute frequently
mentioned Germany as an opponent, but news reports and boycott calls did not focus on
Germany and lacked the vitriol aimed at France. Nonetheless, if American consumers were
switching out of all foreign-sounding brands, they would have boycotted German-sounding
brands, such as Deutschmacher hot dogs, Löwenbräu beer, and Dieffenbachs potato chips.
The results of the German placebo test (Table V, column 1) show American consumers actually increased their purchase of German-sounding products in weeks with more intense
anti-French sentiment. Chavis and Leslie (2009) also found a rise in German wine sales
during this period.
In a related placebo test, we examine whether the dispute prompted consumers to increase their consumption of products that appeared to be imports from prominent US allies
in the Iraq invasion: Great Britain, Spain, and Italy (the “Coalition of the Willing”).21
Although anti-France and pro-Coalition sentiments are not mutually exclusive, evidence of
pro-Coalition buying patterns provides insight into how American consumers conceptualized
their group identity. We construct the variable CoalitionW illingScorei for a given brand i
by summing its perceived language scores for Great Britain, Spain, and Italy and dividing by
three to normalize.22 We estimate our baseline model and include CoalitionW illingScorei
and our main parameter of interest CoalitionW illingScorei ∗ F oxAntiF ranceM entionst .
The model estimates in Table V, column 2 confirm American consumers did not systematically shift into pro-Coalition-sounding brands.
In a variant of this test, we separately consider sales of British-sounding and Italiansounding brands.23 Great Britain was the US’ most steadfast ally throughout the dispute
and most consistently identified as such in public opinion surveys. A March 14 CNN/Gallup
poll found 76% of Americans counted Great Britain among their nation’s allies, whereas only
half of respondents said the same of Spain, the next highest response. We therefore estimate
16
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
the same model but replace the coalition variable with BritishScorei . Examples of highly
British-sounding brands include King Henry’s Pretzels, Parliament Lights cigarettes, and
Royal Cup Coffee. The model estimates in Table V, column 3 show a positive coefficient for
the relevant interaction term but is not statistically different from zero. In a third variant
of this placebo test, we examine the sale of Italian-sounding brands. Among US allies, Italy
has the most precise, readily identifiable mapping of language to perceived country of origin.
Brands survey respondents all deemed Italian include Tassimo coffee, Bellatoria frozen pizza,
and Umberto Giannini shampoo. Model estimates using ItalianScorei are summarized in
Table V, column 4. We find no statistically significant change in those brand sales during
the dispute.
We describe results of four additional robustness tests (see online appendix for full results). First, brands at the higher levels of F renchScorei are on average more expensive,
so F renchScorei may capture aspects of price other than perceived nationality. Gas prices
often fluctuate in response to political conflicts, particularly with regard to military action
in the world’s leading oil-producing region. Consumers paying more for gasoline may have
switched to lower-cost, non-French-sounding brands within the same product category (Ma
et al 2011).24 We add to our baseline model RetailGasolineP ricet the national average
retail price of regular unleaded gasoline in week t and its interaction with perceived French
brand origin. The coefficient for the interaction term RetailGasolineP ricet ∗ F renchScorei
is negative as expected but not statistically significant. This finding further validates our
identifying assumption that the dispute’s influence on consumer behavior was solely through
a negative shock to consumers’ associations with France.
Second, our findings are also robust to controls for popular sentiment regarding the Iraq
invasion that may have correlated with anti-French sentiment. Third, we separately add
to the baseline model the two objective French-origin measures summarized in Table II.
In their respective model interaction, terms F oxAntiF ranceM entionst * F renchHQi and
F oxAntiF ranceM entionst *F renchT rademarki are both negative, indicating a decline in
17
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
sales, but statistically insignificant. These findings confirm objectively French brands saw no
change in market share and underscore the importance of consumer perceptions in explaining
boycott behavior.
Finally, we consider whether the effect of media mentions on purchases levels off by estimating our model with the logged value of media mentions. The log-transformed media measure decreases the importance of weeks with high media activity, making our results less sensitive to outliers. We find the coefficient on F renchScorei ∗Log(F oxN ewsDisputeM entionst )
remains negative and statistically significant, but the coefficient is roughly half that of the
corresponding interaction term in the baseline model. We also lag media variables by one
week to examine the persistence of the dispute’s effects on sales. Although we have strong
theoretical reasons to expect the influence of media mentions dissipates within one week, the
lagged version of the variable captures longer-term effects. The coefficient on the interaction
term F renchScorei ∗ F oxN ewsDisputeM entionst−1 remains negative and significant and is
slightly larger than in the baseline specification.
IV.D
Consumer Demographics as a Source of Cross-sectional Variation
Thus far we have established that during international conflicts, consumers do boycott
products associated with their country’s foreign adversary. In this section, we make an initial
probe of why American consumers varied in the intensity of their response to the US-France
conflict. This store-level analysis correlates market-share fluctuations in a store with the
demographic characteristics of the store’s customers. We analyze cross-sectional variation
in French-sounding brands’ market share in the week ending March 16, 2003, just prior to
the Iraq invasion. As Figure I shows, during that week nearly half of all national survey
respondents were unwilling to purchase French goods.
We hypothesize that international disputes influence consumption decisions by making
national identity more salient in individual decision-making. Specifically, the US-France conflict produced anger toward France, which can heighten national identity salience through a
18
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
variety of mechanisms. In general, consumers choose brands that signal and reinforce their
self-image (Escalas and Bettman 2005; White and Dahl 2006, 2007). External threats to a
country’s national security magnify in/out-group distinctions and motivate behavior to reinforce in-group membership (Huddy and Khatib 2007, Kinder and Kam 2009). International
disputes specifically undermine the psychological foundations of brand loyalty. Consumers
form attachments to specific brands because consistency in consumption goods fosters a subjective sense of stability and safety (Rindfleisch et al 2008). Animosity must be sufficiently
high to supersede pre-existing attachments to French-sounding brands.
We proxy store-level customer demographics using a supplemental IRI dataset of demographic information for a two-mile radius around each supermarket in our sample. This radius is the standard definition of a supermarket’s catchment area. Due to the cross-sectional
nature of these data, we cannot estimate a difference-in-differences model as in the previous
section. Instead, we model the same outcome, weekly change in annual growth share rate
for a single week, as in the previous analysis. This outcome measure holds constant the full
range of time-invariant characteristics, including the ex-ante preference for and availability
of French-sounding brands. Thus, our analysis reveals the correlation between demographics
and the propensity to boycott French-sounding brands.
The IRI demographic data provide two plausible proxies for the strength of customers’ US
national identity: US citizenship and employment in the armed forces. %U SCitizenjt is the
proportion of store j’s customers that are US citizens. Existing studies show citizenship cues
conjure norms of loyalty and allegiance to the state (Theisse-Morse 2009) and that citizens
participate in boycotts more often (Copeland 2014). %U SCitizenjt captures the size of the
population that has either not been in the US long enough to be eligible for citizenship or
are eligible but have chosen not to pursue citizenship. Both groups should have weaker US
allegiances.
Descriptively, stores with higher than average proportions of US citizens more frequently
experience a decline in market share of French-sounding brands. Stores with below aver-
19
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
age proportions of US citizens more frequently see an increase in market share of these
same brands. Of the 116 stores that had higher than average proportions of US citizens
and experienced any change in the store-level average market share of F renchScorei =7
brands, over half (64 stores) experienced a decline. Among the 103 stores with below average proportions of US citizens, more than 60% saw an increase in market share of these highly
French-sounding brands. F renchScorei =3 brands exhibit the same patterns in market-share
change at stores with above and below average numbers of US citizens. For stores with above
average proportions of US citizens, the average market-share change for all F renchScorei ≥
3 brands is -0.00104. Stores with below average proportions of US citizen customers reported
a more modest average market-share change of -0.000324 for the same brands.
Our second measure of consumers’ identification with the US, %ArmedF orcesjt , is the
percent of store j’s customers employed in the US armed forces. We expect that communities
with closer ties to the armed forces should exhibit stronger attachments to US national
identity.25 The months leading up to the 2003 dispute saw extensive US troop deployments
to the Middle East. In our sample, the average store’s customers were only 0.72% armed
forces employees. Although armed forces employment is highly concentrated in our sample,
this variable controls for a subpopulation of consumers most likely to have exceptionally
strong attachments to US national identity. The low correlation between %ArmedF orcesjt
and %U SCitizenjt confirms the two variables are distinct.
Partisanship is another plausible source of variation in boycott participation, although its
effects are ambiguous. Fox News’ prominent role during this conflict suggests Republicans
were more likely to change consumption in response to the conflict. Although we cannot
verify whether Fox News coverage caused boycott responses, Fox News viewers may have
been ex ante more likely to boycott regardless of news coverage tone.
Most political behavior research, however, concludes partisanship is a weak predictor
of sentiment. Ethnocentrism, a way of seeing the world in terms of us and them, has
been shown to be a better predictor of support for homeland security spending, military
20
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
intervention, and opposition to foreign aid in the wake of the September 11th terrorist attacks
than partisanship (Kinder and Kam 2009). Conservatives tend to express ethnocentrism
more often than liberals (Huddy 2013), but conservatives have stickier supermarket brand
preferences owing to a greater aversion to change (Khan et al 2013). Copeland (2014) finds
US survey respondents with more intense partisan identities report lower participation in
politically-motivated boycotts.
We test this alternate explanation by including %BushV otejt , George W. Bush’s popular
vote share in the 2000 presidential election in the county in which store j is located. We
construct this measure from county-level election returns reported in David Liep’s Atlas of
US Presidential Elections.26 Unfortunately election returns data are available at only the
county level, which makes for a coarser measure than our store-level demographic data. The
low correlation between %U SCitizenjt and %BushV otejt (0.33) indicates citizenship is not
merely a proxy for partisanship.
Finally, we control for household income in case consumers’ price sensitivity influences
their propensity to boycott. M edianHouseholdIncomejt is the median household income of
store j’s customers. The correlation between %U SCitizenjt and M edianHouseholdIncomejt
is -0.11. As in the baseline model in the previous section, we control for year-over-year change
in price and number of varieties stocked for the previous week and include the one-week
lagged dependent variable. These three variables control for the main sources of exogenous
change in brand market-share growth rate.
Given that demographics are fixed over our time period of interest, we model crosssectional variation as a function of each store’s customer demographics. Specifically, we
estimate weekly change in annual market-share growth rate for each brand-product categorystore for the week ending March 16:
∆Share03 -02ijkt = β1 %U SCitizenjt +β2 M edianHouseholdIncomejt +β3 %ArmedF orcesjt +
β4 %BushV otejt + β5 F renchScorei + β6 %U SCitizenjt ∗ F renchScorei +
β7 M edianHouseholdIncomejt ∗F renchScorei +β8 %ArmedF orcesjt ∗F renchScorei +
21
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
β9 %BushV otejt ∗ F renchScorei + β10 ∆Price03 -02ijkt +
β11 ∆NumVariants03 -02ijkt + β12 ∆Share03 -02ijkt−1 + ijkt−1
where t = the week ending March 16, 2003.
Our main coefficient of interest, β6 , indicates the degree a higher proportion of US citizens
corresponds to sharper decline in the market share of French-sounding brands. A significant
coefficient would indicate that in the week ending March 16, stores with more US citizens
demonstrate a stronger boycott response. Table VI summarizes our model estimates.
We find sales of French-sounding brands declined in communities with higher proportions
of US citizens. Comparing column 1, the simplest specification, and column 5, including the
full set of controls, the β6 point estimate is slightly higher in the fully specified model. These
results show that for every percentage point increase in the proportion of US citizens, the
average store experienced a 0.2 percent point drop in French-sounding brands’ market share
compared to the prior week. Comparing stores at the mean value of %U SCitizenjt to stores
that are one standard deviation above the mean, the latter group had on average a 1.4%
larger drop in French-sounding brands’ market share.
The other demographic variables and their associated interactions with F renchScorei
do not achieve conventional levels of statistical significance in any specification. The partisanship interaction term has a positive sign but inferences are limited not only because of
the insignificant point estimate but also because it is measured at a higher level of aggregation than other variables. The three controls for the main drivers of market-share change,
lagged share change, price change, and change in the number of brand varieties, are highly
significant.
In a robustness test, we address the possibility that unobserved, time-invariant correlates
of customer traits drive observed market-share fluctuations. If that were true, the week
ending March 16 is not particularly salient and we cannot interpret demographic traits
as correlates of boycott response. For example, given that armed forces employment is
concentrated in one region of Virginia, we need to exclude the possibility that some other
22
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
unobserved characteristic of this region does not drive our findings. We test this alternative
explanation by estimating the full model with a pooled sample of all other weeks in 2003.
The point estimate for %U SCitizenjt ∗ F renchScorei is not statistically significant in this
sample, indicating citizenship was particularly salient at the peak of the US-France crisis.
V
Conclusion
International conflicts often provoke boycott calls targeted at a foreign adversary, but
whether consumers actually participate has been an enduring puzzle. In this article, we
show that US consumers reduced their purchases of French-sounding supermarket brands
during the 2003 US-France dispute over the Iraq invasion. We show that at the dispute’s
peak, supermarkets with a higher proportion of US citizens in their customer base saw a
sharper decline in sales of French-sounding brands. This finding suggests animosity toward
a foreign adversary taps into consumers’ national identity.
Amid deeper global economic integration, the influence of international political conflicts on consumer behavior is poised to grow. The specter of large-scale military conflicts
that loomed over much of the 20th century has given way to small-scale but frequent disputes. Increasingly, these conflicts pit advanced market economies against emerging market
countries. The size and growing prosperity of these markets make them crucial for global
producers of consumer goods. Recurring political conflicts may damage the reputation of
foreign nationality-based brands among emerging market consumers (Deshpande 2010, Chu
2013).
This study also highlights the untapped potential of consumer behavior as a metric of
popular sentiment towards countries. We analyze weekly supermarket sales data, but related
metrics, including individual- and household-level panel data, hold the promise of even more
nuanced studies that link preferences over time to individual and community characteristics.
The influence of politics on consumption behavior is particularly timely in the US. Recent
changes in US campaign finance laws give private corporations unprecedented political influ23
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
ence. Amid heated discussion on the public disclosure of these activities, consumer behavior
is likely to become more important venue for political expression. Our findings reveal a risk
to corporate political activity in the form of consumer backlash and highlight a mechanism
by which consumers can hold corporate donors accountable. These developments portend
an even tighter connection between consumer behavior and politics.
Notes
1
Following the 2003 Iraq invasion, anti-American sentiment in several Middle Eastern
countries fueled calls to boycott US products (Clerides et al 2011). In 2005, efforts to boycott
Danish products emerged in those same countries following the publication of anti-Islam
cartoons in a Danish newspaper (Maher and Mady 2010). In the 2000s, China’s diplomatic
conflicts with France and Japan stoked boycott calls against both countries (Hong et al 2011,
Fisman et al 2013).
2
Sources of ex ante brand preferences include product characteristics like price and quality
(Zeithaml 1988), early life exposure (Bronnenberg et al 2012), advertising (Bagwell 2007),
and peer influence (Villanueva et al 2008). Brand choice is also an expression of consumers’
self-image including their political ideology (Khan et al 2013), self-worth (Shachar et al
2011), and social beliefs and values (Muniz and O’Guinn 2001).
3
Studies specifically show that American consumers overlook legally mandated country of
origin product labels (Samiee et al 2005). The only legal prohibition on branding/advertising
that implies foreign origin regards illegal use of protected geographic indicators (e.g. Champagne, Cognac).
4
There was no risk that the US-France dispute or the anticipated Iraq invasion would
systematically influence the market supply of French-sounding brands. For each store, we
restrict our sample to brands that were sold throughout 2002 and 2003 in case brands were
introduced or removed for reasons related to the conflict. In the short term, stores are
contractually barred from pulling or relocating brands from their shelves.
24
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
5
At this point, the US and UK withdrew the Security Council resolution so no actual
vote took place.
6
Ashenfelter, Ciccarella, and Shatz (2007) report the first mentions of “freedom fries” in
print and television news were February 20 and February 17 respectively.
7
IRI set its market definitions in 1987 to achieve a representative sample of US consumers
making it unlikely that our findings are artifact of sample selection. See online appendix for
a list of IRI geographic market names.
8
We performed a trial experiment to test whether additional brand information influ-
enced perceived nationality. For a random sample of brands with US-trademarked logos, we
surveyed a randomly selected group on the nationality of brands based on the brand name,
product category, and logo. A control group scored the same brands based only on brand
name and product category. Answers were not statistically distinguishable between the two
groups.
9
10
See online appendix (Figure A2) for a sample US trademark filing.
For example, Dannon is a French company but the trademark is registered to Dannon’s
US subsidiary. Yoplait is a French-owned trademark licensed to the American company
General Mills. Trade data does not detect either form of French product nationality.
11
These brands include L’ORÉAL, L’ORÉAL COLOR VIVE (shampoo); PIERRE CARDIN
(deodorant); BORNIER, POMMERY, TEMERAIRE (mustard/ketchup); FISCHER LA
BELLE AMBER, FISCHER LA BELLE, DESPERADOS, BRASSEURS (beer); YOPLAIT
KIDS, YOPLAIT GO GURT (yogurt).
12
The other brands are MATRIX AMPLIFY, REDKEN, three varieties of DANNON
CHUNKY, and YOPLAIT CARB MONITOR.
13
See online appendix for a detailed description of coding procedures.
14
DellaVigna and Kaplan (2007) exploit the randomized expansion of cable television
access to show that exposure to Fox News causes an increase in conservative sentiment.
15
We verify weekly price changes are uncorrelated with brands’ F renchScorei .
25
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
16
Manufacturers provide retailers with a trade allowance to finance price promotions.
17
Manufactures negotiate with retailers for specific shelf locations for their products. Local
distributors stock shelves and can monitor compliance. These agreements are negotiated
chain-wide and renegotiated at fixed intervals.
18
As is standard in marketing research we also evaluated a model with lagged price. Cur-
rent price may be problematic because there are likely to be common shocks that affect price
and sales simultaneously (e.g., conflict with France causes both declining demand and in response, a drop in price to counteract declining demand). Our findings are however robust
to the use of lagged price.
19
The figures also indicate brand-share growth returns to baseline levels relatively quickly.
With store-level data, we are unable to say whether individuals temporarily switched out of
French brands or if the switch was long-term but compensated for by other consumers.
20
Using our baseline model estimates, we calculate the cumulative cost in lost sales of
F renchScorei ≥ 3 brands for that week. We estimate the average per-brand loss based on
sample averages for a brand’s change in market share, total units sold, and price.
21
The coalition had 40-50 countries at any given time. We focus on the three listed
because they were prominent European allies whose support was in direct contrast to French
and German opposition. Additionally, these countries are well represented in our perceived
nationality scores whereas other coalition members appear only sporadically.
22
The measure is somewhat noisy in that English and Spanish linguistic cues can corre-
spond to multiple countries, but this noise mimics the true ambiguity consumers face.
23
We do not perform an analogous test for Spanish-sounding brands. While Spain sup-
ported the invasion, other salient Spanish-speaking countries like Mexico opposed it.
24
Any rise in gas prices in anticipation of the Iraq invasion would prompt greater overall
consumption of grocery products because price-conscious consumers purchase more groceries
and reduce their consumption of restaurant meals (Gicheva, Hastings, and Villas-Boas 2010).
25
Arguably, even civilian armed forces employees would have a heightened sense of US
26
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
national identity.
26
These data are widely used, including in Khan et al 2013, which analyzes partisanship
and brand preferences using the same IRI data we use in the previous section.
References
[1] Abraham, Magid M. and Leonard M. Lodish, “Getting the Most Out of Advertising and
Promotion,” Harvard Business Review, May-June (1990), 50-60.
[2] Ashenfelter, Orley, Stephen Ciccarella, and Howard J. Shatz. “French Wine and the U.S.
Boycott of 2003: Does Politics Really Affect Commerce?,” Journal of Wine Economics 2,
no. 1 (2007), 55-74.
[3] Assmus, Gert, John U. Farley, and Donald R. Lehmann. “How Advertising Affects Sales:
Meta-analysis of Econometric Results.” Journal of Marketing Research (1984): 65-74.
[4] Ataman, M. Berk, Harald J Van Heerde, and Carl F. Mela. “The Long-Term Effect of
Marketing Strategy on Brand Sales,” Journal of Marketing Research 47, no. 5 (2010),
866-882.
[5] Bagwell, Kyle. “The Economic Analysis of Advertising,” in Handbook of Industrial Organization Vol. 3, Mark Armstrong and Robert Porter eds. (North-Holland: Amsterdam
2007, 1701-1844).
[6] Balabanis, George, and Adamantios Diamantopoulos. “Gains and Losses from the Misperception of Brand Origin: The Role of Brand Strength and Country-of-Origin Image,”
Journal of International Marketing 19, no. 2 (2011), 95-116.
[7] Baum, Matthew, and Tim J. Groeling. War Stories: The Causes and Consequences of
Public Views of War. (Princeton, NJ: Princeton University Press, 2010).
[8] Bordalo, Pedro, Nicola Gennaioli, and Andrei Shleifer. “Salience and Consumer Choice,”
Journal of Political Economy 121.5 (2013): 803-843.
27
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
[9] Bronnenberg, Bart J., Jean-Pierre Dubé, and Matthew Gentzkow. “The Evolution of
Brand Preferences: A Consumer Migration Approach,” American Economic Review 102,
no. 6 (2012), 2472-2508.
[10] Bronnenberg, Bart J., Michael W. Kruger, Carl F. Mela. “Database paper: The IRI
Marketing Data Set,” Marketing Science 27, no. 4 (2008), 745-748.
[11] Chavis, Larry, and Phillip Leslie. “Consumer Boycotts: The Impact of the Iraq War on
French Wine Sales in the U.S.,” Quantitative Marketing and Economics 7 (2009), 37-67.
[12] Chu, Junhong. “Quantifying Nation Equity with Sales Data: A Structural Approach,”
International Journal of Research in Marketing 30 (2013), 19-35.
[13] Clerides, Sofronis, Peter J. Davis, and Antonis Michis. “The Impact of the Iraq War on
U.S. Consumer Goods Sales in Arab Countries,” SSRN eLibrary (2011). Accessed August
14, 2012: http://papers.ssrn.com/sol3/papers.cfm?abstract id=1687180
[14] Copeland, Lauren. “Conceptualizing Political Consumerism: How Citizenship Norms
Differentiate Boycotting from Buycotting,” Political Studies 62, no. S1 (2014), 172-186.
[15] Davis, Christina and Sophie Meunier. “Business as Usual: Economic Responses to Political Tensions,” American Journal of Political Science 55, no. 3 (2011), 628-646.
[16] DellaVigna, Stefano, and Ethan Kaplan. “The Fox News Effect: Media Bias and Voting,” Quarterly Journal of Economics 122, no. 3 (2007), 1187-1234.
[17] Deshpande, Rohit. “Why You Aren’t Buying Venezuelan Chocolate,” Harvard Business
Review, December (2010).
[18] Escalas, Jennifer Edson and James Bettman. “Self-Construal, Reference Groups, and
Brand Meaning,” Journal of Consumer Research 32, no. 3 (2005), 378-89.
28
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
[19] Fisman, Raymond, Yasushi Hamao, and Yongxiang Wang. “The Impact of Interstate
Tensions on Economic Exchange: Evidence from Shocks to Sino-Japanese Relations.”
Unpublished working paper. Columbia University, May (2013).
[20] Foulka, Vasiliki and Hans-Joachim Voth. “Reprisals Remembered: German-Greek Conflict and Car Sales During the Euro Crisis,” Unpublished working paper. Universitat
Pompeu Fabra, May (2013).
[21] Fuchs, Andreas, and Nils-Hendrik Klann. “Paying a Visit: The Dalai Lama Effect on
International Trade,” Journal of International Economics 91 no. 1 (2013), 164-177.
[22] Gennaioli, Nicola and Andrei Shleifer. “What Comes to Mind.” Quarterly Journal of
Economics 75, no. 4 (2010), 1399-1433.
[23] Gentzkow, Matthew, and Jesse M. Shapiro. “Media Bias and Reputation,” Journal of
Political Economy 114, no. 2 (2006), 280-316.
[24] Gerber, Alan S., James G. Gimpel, Donald P. Green, and Daron R. Shaw. “How Large
and Long-lasting Are the Persuasive Effects of Televised Campaign Ads? Results from
a Randomized Field Experiment,” American Political Science Review 105, no. 1 (2011),
135-150.
[25] Gicheva, Dora, Justine Hastings, and Sofia Villas-Boas. “Investigating Income Effects
in Scanner Data: Do Gasoline Prices Affect Grocery Purchases?,” American Economic
Review 100, no. 2 (2010), 480-484.
[26] Glick, Reuven, and Alan M. Taylor. “Collateral Damage: Trade Disruption and the
Economic Impact of War,” Review of Economics and Statistics 92, no. 1 (2010), 102-27.
[27] Groseclose, Tim, and Jeffrey Milyo. “A Measure of Media Bias,” Quarterly Journal of
Economics 120, no. 4 (2005), 1191-1237.
29
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
[28] Hong Canhui, Wei-Min Hu, James Preiger, Dongming Zhu. “French Automobiles and
the Chinese Boycotts of 2008: Politics Really Does Affect Commerce,” The B.E. Journal
of Economic Analysis & Policy 11:1 (2011).
[29] Huddy, Leonie. “Translating Group Identity into Political Cohesion and Commitment,”
In Oxford Handbook of Political Psychology. Leonie Huddy, David O. Sears, and Jack Levy
(Eds.). New York: Oxford University Press. 2nd. Edition (2013).
[30] Huddy, Leonie, and Nadia Khatib. “American Patriotism, National Identity, and Political Involvement,” American Journal of Political Science 51, no. 1 (2007), 63-77.
[31] Iyengar, Shanto, and Kyu S Hahn. “Red Media, Blue Media: Evidence of Ideological
Selectivity in Media Use,” Journal of Communication 59, no. 1 (2009), 19-39.
[32] John, Andrew, and Jill Klein. “The Boycott Puzzle: Consumer Motivations for Purchase
Sacrifice,” Management Science 49, no. 9 (2003), 1196-1209.
[33] Khan, Romana, Kanishka Misra, and Vishal Singh. “Ideology and Brand Consumption,”
Psychological Science 24, no. 3 (2013), 326-33.
[34] Kinder, Donald R., and Cindy D. Kam. Us Against Them: Ethnocentric Foundations
of American Opinion. (Chicago: University of Chicago Press, 2009).
[35] Klein, Jill Gabrielle. “Us versus Them, or Us versus Everyone? Delineating Consumer
Aversion to Foreign Goods,” Journal of International Business Studies 33, no. 2 (2002),
345-363.
[36] Kőszegi, Botond and Adam Szeidl. “A Model of Focusing in Economic Choice,” Quarterly Journal of Economics 128, no. 1 (2013), 53-104.
[37] Leclerc, France, Bernd H. Schmitt, and Laurette Dubé. “Foreign Branding and Its
Effects on Product Perceptions and Attitudes,” Journal of Marketing Research 31, no. 2
(1994), 263-270.
30
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
[38] Liu, Scott S., and Keith F. Johnson. “The Automatic Country-of-Origin Effects on
Brand Judgments,” Journal of Advertising 34, no. 1 (2005), 87-97.
[39] Maher, Amro A., and Sarah Mady. “Animosity, Subjective Norms, and Anticipated
Emotions During an International Crisis,” International Marketing Review 27, no. 6
(2010), 630-651.
[40] Ma, Yu, Kusum L. Ailawadi, Dinesh K. Gauri, and Dhruv Grewal. “An Empirical
Investigation of the Impact of Gasoline Prices on Grocery Shopping Behavior,” Journal
of Marketing 75, no. 2 (2011), 18-35.
[41] Martin, B., Lee, M. S. W., and Lacey, C. “Countering Negative Country of Origin Effects
using Imagery Processing,” Journal of Consumer Behaviour 10, no. 2 (2011), 80-92.
[42] Michaels, Guy, and Xiaojia Zhi. “Freedom Fries,” American Economic Journal: Applied
Economics 2 (2010), 256281.
[43] Mitchell, Dona-Gene. “It’s about Time: The Lifespan of Information Effects in a Multiweek Campaign,” American Journal of Political Science 51, no. 1 (2012), 271-518.
[44] Morwitz, Vicki G., and David Schmittlein. “Using Segmentation to Improve Sales Forecasts Based on Purchase Intent: Which ‘Intenders’ Actually Buy?,” Journal of Marketing
Research 29, no. 4 (1992), 391-405.
[45] Muniz, Albert M. and Thomas C. O’Guinn. “Brand Community,” Journal of Consumer
Research, 27 (March 2001), 412-32.
[46] Niederdeppe, Jeff, and Dominick L. Frosch. “News Coverage and Sales of Products with
Trans Fat,” American Journal of Preventive Medicine 36, no. 5 (2009), 395-401.
[47] Rindfleisch, Aric, James E. Burroghs and Nancy Wong. “The Safety of Objects: Materialism, Existential Insecurity, and Brand Connection,” Journal of Consumer Research
36 (June 2008), 1-16.
31
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
[48] Russell, Cristel Antonia. “Investigating the Effectiveness of Product Placements in Television Shows: The Role of Modality and Plot Connection Congruence on Brand Memory
and Attitude,” Journal of Consumer Research 29, no. 3 (2002), 306-318.
[49] Sen, Sankar , Zeynep Gürhan-Canli, and Vicki Morwitz. “Withholding Consumption:
A Social Dilemma Perspective on Consumer Boycotts,” Journal of Consumer Research
28, no. 3 (2001), 399-417.
[50] Samiee, Saeed, Terence A. Shimp, and Subhash Sharma. “Brand Origin Recognition Accuracy: Its Antecedents and Consumers’ Cognitive Limitations,” Journal of International
Business Studies 36, no. 4 (2005), 379-397.
[51] Shachar, Ron, Tülin Erdem, Keisha M. Cutright, and Gavan J. Fitzsimons. “Brands:
The Opiate of the Nonreligious Masses?,” Marketing Science 30, no. 1 (2011), 92-110.
[52] Theiss-Morse, Elizabeth. Who Counts as an American?: The Boundaries of National
Identity. New York, Cambridge University Press (2009).
[53] Usunier, Jean-Claude, and Janet Shaner. “Using Linguistics for Creating Better International Brand Names,” Journal of Marketing Communications 8 (2002), 211-228.
[54] Villanueva, J., S. Yoo, and D.M. Hanssens. “The Impact of Marketing-Induces vs Wordof-Mouth Customer Acquisition on Customer Equity,” Journal of Marketing 45, no. 1
(2008), 48-59.
[55] White, Katherine and Darren W. Dahl. “To Be or Not Be? The Influence of Dissociative
Reference Groups on Consumer Preferences,” Journal of Consumer Psychology 16, no. 4
(2006), 404-413.
[56] White, Katherine and Darren W. Dahl. “Are All Out-Groups Created Equal? Consumer
Identity and Dissociative Influence,” Journal of Consumer Research 34 (December 2007),
525-36.
32
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
[57] Yang, Sha, Yuxin Chen, and Greg M. Allenby. “Bayesian Analysis of Simultaneous
Demand and Supply.” Quantitative Marketing and Economics 1 no. 3 (2003): 251-275.
[58] Zeithaml, Valarie A. “Consumer Perceptions of Price, Quality, and Value: A Means-End
Model and Synthesis of Evidence,” Journal of Marketing 52, no. 3 (1988).
Department of Politics, University of Virginia, Charlottesville, VA
Darden School of Business, University of Virginia, Charlottesville, VA
33
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
Table I: Brand Examples across F renchScorei Values
F renchScorei Values
7
Brand Example (Product Category)
L’oréal (shampoo)
Sans Sucre de Paris (sugar substitute)
6
Grey Poupon (mustard/ketchup)
Elliare (facial tissue)
5
Cartier Vendôme Lights (cigarettes)
Dogfish Head Raison D’Être (beer)
4
Hormel Dubuque (hot dog)
Piraquê (salty snacks)
3
French’s Mustard (mustard/ketchup)
Pinol (household cleaning)
2
Pasquale’s (frozen pizza)
Blue Bonnet (margarine/butter)
1
Stouffer’s Lean Cuisine (frozen dinner)
Armour (hot dog)
0
Budweiser (beer)
Sensodyne (toothpaste)
F renchScorei =Number of survey respondents that deem brand i to be French
34
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
Table II: Distribution of Brands across Perceived and Actual French Origin
Perceived
Actual
French
F renchScorei
0
Headquarters1
7
Non-French-owned
French-owned
Not
US Trademark2
3502
US Trademark2
7
Trademarked2
1208
Total
4717
1
12
1723
1
484
2208
2
1
690
1
177
868
3
18
355
3
74
432
4
3
161
3
29
193
5
4
114
5
36
155
6
2
29
4
1
34
7
4
23
3
11
37
Most brands consumers perceive as French are not actually owned by French corporations.
Few brands in our sample are imported from France or owned by French corporations.
However, those that are French-owned are likely to be supplied by corporations
headquartered in France. Brand and product category classification are from supermarket
scanner data supplied by IRI Inc.
1
Source: author’s calculations based on Thomson Reuters Worldscope and Lexis-Nexis
databases
2
Source: author’s calculations based on US Patent and Trademark Office database
35
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
Table III: Weekly Change in French Brands’ Share Growth: Baseline Model
F renchScorei
(1)
(2)
5.94E-5
3.41E-5
(1.98E-5)
(1.98E-5)
F oxT otalDisputeM entionst
4.37E-6
(3.04E-6)
F oxT otalDisputeM entionst ∗ F renchScorei
-1.35E-5
(3.59E-6)
F oxAntiF ranceM entionst
1.25E-5
(6.33E-6)
F oxAntiF ranceM entionst ∗ F renchScorei
-4.88E-5
(7.44E-6)
∆Price03 -02ijkt
∆NumVariants03 -02ijkt
∆Share03 -02ijkt−1
R2
Observations
-1.19E-2
-1.19E-2
(4.35E-5)
(4.34E-5)
8.9E-1
8.92E-1
(4.31E-4)
(4.31E-4)
2.92E-2
2.92E-2
(1.91E-4)
(1.91E-4)
0.1592
0.1592
23,199,220
23,199,220
Standard errors in parentheses; coefficients significant at (two-tailed) p < 0.01 are in bold.
36
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
Table IV: Anti-French vs. Pro-French Sentiment
F renchScorei
F oxDisputeM entionst
(1)
(2)
2.9E-5
5.09E-5
(2.02E-5)
(2.00E-5)
4.41E-6
(3.59E-6)
F oxDisputeM entionst ∗ F renchScorei
-1.62E-5
(4.23E-6)
CN N DisputeM entionst
-8.13E-8
(3.22E-6)
CN N DisputeM entionst ∗ F renchScorei
4.55E-6
(3.79E-6)
F oxAntiF ranceM entionst
1.74E-5
(7.19E-6)
F oxAntiF ranceM entionst ∗ F renchScorei
-6.12E-5
(8.45E-6)
CN N P roF ranceM entionst
-9.67E-6
(6.77E-6)
CN N P roF ranceM entionst ∗ F renchScorei
2.48E-5
(7.98E-6)
∆Price03 -02ijkt
37
-1.19E-2
-1.19E-2
(4.34E-5)
(4.35E-5)
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
∆NumVariants03 -02ijkt
8.92E-1
8.92E-1
(4.31E-4)
(4.31E-4)
2.92E-2
2.92E-2
(1.91E-4)
(1.91E-4)
0.1592
0.1592
23,199,220
23,199,220
∆Share03 -02ijkt−1
R2
Observations
Standard errors in parentheses; coefficients significant at (two-tailed) p < 0.01 are in bold.
38
39
BritishScorei
F oxAntiF ranceM entionst ∗ CoalitionW illingScorei
CoalitionW illingScorei
F oxAntiF ranceM entionst ∗ GermanScorei
GermanScorei
F oxAntiF ranceM entionst
(6.26E-6)
4.47E-5
(1.66E-5)
(1.32E-5)
-8.84E-6
(3.49E-5)
2.57E-5
(1.27E-5)
(6.40E-6)
-5.32E5
6.97E-6
(2)
-1.58E-5
(1)
(3)
(1.12E-5)
-8.56E-7
(1.16E-5)
-8.83E-6
Table V: Placebo Tests Based on Other Nationalities
(1.88E-5)
1.33E-6
(4)
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
40
23,206,389
23,206,389
0.1609
(2.32E-4)
(1.91E-4)
0.1592
2.92E-2
(5.03E-4)
(4.31E-4)
2.92E-2
8.60E-1
(6.09E-5)
-1.46E-2
8.92E-1
(4.34E-5)
-1.19E-2
23,206,389
.1609
(2.32E-4)
2.92E-2
(5.03E-4)
8.60E-1
(6.09E-5)
-1.46E-2
(4.00E-6)
5.21E-6
Standard errors in parentheses; coefficients significant at (two-tailed) p < 0.01 are in bold.
Observations
R2
∆Share03 -02ijkt−1
∆NumVariants03 -02ijkt
∆Price03 -02ijkt
F oxAntiF ranceM entionst ∗ ItalianScorei
ItalianScorei
F oxAntiF ranceM entionst ∗ BritishScorei
23,206,389
.1609
(2.32E-4)
2.92E-2
(5.03E-4)
8.60E-1
(6.09E-5)
-1.46E-2
(7.15E-6)
-2.81E-6
(1.88E-5)
-1.48E-7
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
41
M edianHouseholdIncomejt
%BushV otejt ∗ F renchScorei
%BushV otejt
%U SCitizenjt ∗ F renchScorei
%U SCitizenjt
F renchScorei
(4.52E-3)
(7.35E-4)
-1.48E-3
1.60E-4
(3.48E-4)
(3.43E-4)
(1.11E-3)
(1.04E-3)
1.15E-4
1.31E-3
(9.51E-4)
(8.91E-4)
6.37E-4
3.03E-4
5.14E-4
(7.81E-4)
-1.92E-3
(6.83E-4)
(1.30E-4)
6.07E-3
(5)
(6.41E-4)
(4.33E-3)
-6.13E-4
(4)
6.13E-4
(4.99E-4)
(6.07E-4)
5.09E-3
(3)
6.29E-4
-8.88E-4
(2)
6.05E-4
(1)
Table VI: Consumer Demographics and Cross-Sectional Variation
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
42
468,811
468,811
.1639
(1.34E-3)
-8.85E-2
(3.08E-3)
8.96E-1
(3.11E-4)
-1.04E-2
468,811
.1640
(1.33E-3)
-8.85E-2
(3.08E-3)
8.95E-1
(3.11E-4)
468,81
.1639
(1.34E-3)
-8.85E-2
(3.08E-3)
8.96E-1
(3.11E-4)
-1.07E-2
(4.39E-3)
(4.37E-3)
-1.06E-2
3.31E-3
(3.57E-3)
(3.56E-3)
2.84E-3
-1.79E-3
(4.07E-4)
(4.02E-4)
-1.55E-3
-5.33E-4
-5.28E-4
Standard errors in parentheses; coefficients significant at (two-tailed) p < 0.05 are in bold. t = the week ending March 16,
2003
468,811
Observations
0.1639
(1.34E-3)
(1.34E-3)
0.1639
-8.85E-2
(3.08E-3)
(3.08E-3)
-8.85E-2
8.96E-1
(3.11E-4)
(3.11E-4)
8.96E-1
-1.07E-2
-1.07E-2
R2
∆Share03 -02ijkt−1
∆NumVariants03 -02ijkt
∆Price03 -02ijkt
%ArmedF orcesjt ∗ F renchScorei
%ArmedF orcesjt
M edianHouseholdIncomejt ∗ F renchScorei
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
43
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
Figure I: 2003 US Sentiment toward France and Iraq War
This figure plots three dimensions of public opinion related to the 2003 US-France conflict:
general support for the Iraq War, views towards France, and willingness to purchase French
imports. Points for the latter two correspond to weeks in which public opinion were
available. Data on Iraq War attitudes are available for all weeks. All values are mean US
public opinion for that week, averaging across all polls in the Roper Center’s iPoll
Database that posed the question. See online appendix for a list of polls and the questions
used to construct average opinion. The plot highlights key dates in the timeline of the
conflict. The green dashed line indicates the week in which the invasion began. In the week
ending February 16, the French foreign minister spoke before the UN to argue against the
proposed invasion. In the week prior, US Secretary of State Colin Powell presented to the
UN supposed evidence of Iraqi weapons programs. In the week ending March 16, the US
House of Representative renamed French fries “freedom fries.” In the week prior, the US
and France ended negotiations to continue weapons inspections in lieu of an invasion.
44
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
Figure II: Weekly News Coverage of French Boycott
This figure plots the total weekly news programs that mention the French boycott on Fox
News (top panel) and CNN (bottom panel). Total counts include neutral mentions of the
boycott.
Source: Lexis-Nexis.
45
46
Market boundaries are defined by IRI Inc. Colors distinguish adjacent markets. See online appendix for list of market names.
Figure III: Geographic Coverage of Weekly Supermarket Sales Data
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+
-
47
French-sounding brands.
anti-French sentiment. The most French-sounding brands saw sharper declines in market share than moderately
given week. The figure illustrates that the declining market share of French-sounding brands corresponds with spikes in
(F renchScorei = 3). The dotted line indicates the number of times the US-France dispute was mentioned on Fox News in a
(F renchScorei = 7). The gray line plots the same variable for moderately French-sounding brands in the sample
The black line plots the weekly fluctuation in annual brand share growth of the most French-sounding brands in the sample
Figure IV: Market Share of French-Sounding Brands Declines When Anti-French Sentiment Greatest
526
!"
##$
%%&'(")"*%%
+
# ,#
'
#%+